Electrical audio signal processing systems and devices – One-way audio signal program distribution – Public address system
Patent
1990-01-04
1992-03-31
Shaw, Dale M.
Electrical audio signal processing systems and devices
One-way audio signal program distribution
Public address system
G10L 708
Patent
active
051014343
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
The present invention relates to a method and apparatus for speech recognition.
Voice recognition systems are known. However, such systems, which operate on the principle of dividing the sounds into frequency bands by means of filters and then analysing the energy levels in each band, are relatively expensive. Furthermore, isolated word recognition systems based upon Time Encoded Speech (TES) which do not rely upon the principle of dividing the sounds into frequency bands, are also known.
A system and procedure for isolated word recognition using Time Encoded Speech is described in "Verification, Archetype Updating, and Automatic Token Set Selection, as a means of improving the performance of Menu Driven Isolated Word Recognition Systems using Time Encoded Speech Descriptors in High Acoustic Noise Backgrounds" by R. C. Power, R. D. Hughes and R. A. King; proceedings of International Conference Speech Input/Output Techniques and Applications (1986) pp 144-151.
TES is a form of speech waveform coding. The speech waveform is broken into time intervals (epochs) between sucessive real zeros. For each epoch of the waveform the code consists of a single digital word. This word is derived from two parameters of the epoch, its quantized time duration and its shape. The measure of duration is straightforward, and the commonly adopted strategy for shape description is to classify epochs on the basis of the number of positive minima or negative maxima occurring therein. For economical coding the number of naturally occurring distinguishable symbols produced by this process may then be mapped in a non-linear fashion onto a much smaller number (alphabet) of code descriptors. An algorithm to perform an initial TES coding is described in "Time Encoded Speech (TES) Descriptors as a Symbol Feature Set for Voice Recognition Systems, by J. Holbeche, R. D. Hughes and R. A. King, Proceedings of The International Conference Speech Input/Output Techniques and Applications (1986) pp 310-315.
Isolated word recognition systems based upon TES have many advantages over frequency division systems and are particularly advantageous in high ambient noise environments. However, such systems sometimes exhibit limitations in their ability to cope with connected or continuous word recognition tasks.
SUMMARY OF THE INVENTION
It will be appreciated, therefore, that there is a need for an improved voice recognition system based upon time encoded speech (TES) to cope with connected or continuous recognition tasks. It is an object of the present invention to provide an improved method and apparatus for recognising voice signals, and in particular, voice signals encoded as time encoded speech.
Accordingly, there is provided a method for recognizing voice signals encoded as time encoded speech (TES), the method comprising segmenting a time encoded speech symbol stream into a number of time frames and applying each time frame to a plurality of seeker circuits, each seeker circuit being optimized to detect an acoustic event of the voice signals to be recognized, examining parameters of TES symbols in the time frames thereby to determine the presence or absence of any acoustic event to which any seeker circuit is optimized, determining segmentation boundaries in the TES symbol stream in dependence upon the acoustic events determined as present or absent, comparing parameters of the TES symbol stream within the segmentation boundaries with archetypes of words or utterances stored as time encoded speech thereby to provide an output signal indicative of the nature of the voice signal.
In accordance with the present invention, there is also provided an apparatus for recognizing voice signals encoded as time encoded speech. The apparatus comprises receiver means for receiving a time encoded speech symbol stream and segmenting the stream into a number of time frames, a plurality of seeker circuits, arranged to receive the time frames and being optimized for detecting an acoustic event of the voice signals to be recognized, cla
REFERENCES:
patent: 3679830 (1972-07-01), Uffelman et al.
patent: 4763278 (1988-08-01), Rajaskaran et al.
patent: 4783807 (1988-11-01), Marley
patent: 4852170 (1989-07-01), Bordeaux
Scarr, "Work Recognition Machine", Proc IEEE, vol. 117, No. 1, 1/70, pp. 203-212.
Itahashi, "Discrete Word Recognition Utilizing a Word Dictionary", IEEE Trans. Audio and Elect., vol. Au 21, No. 3, 8/73, pp. 239-248.
Doerrler Michelle
Shaw Dale M.
LandOfFree
Voice recognition using segmented time encoded speech does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Voice recognition using segmented time encoded speech, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Voice recognition using segmented time encoded speech will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2265083